End-to-End Quality of Service over Cellular...

End-to-End Quality of Service over Cellular Networks Data Services Performance and Optimization in 2G/3G

Edited by

G. Gmez and R. Snchez Both ofOptimi Corporation Spain

ffirs.fm Page iii Tuesday, March 22, 2005 7:26 AM

Innodata0470015861.jpg

ffirs.fm Page ii Tuesday, March 22, 2005 7:26 AM

End-to-End Quality of Service over Cellular Networks

ffirs.fm Page i Tuesday, March 22, 2005 7:26 AM

ffirs.fm Page ii Tuesday, March 22, 2005 7:26 AM

End-to-End Quality of Service over Cellular Networks Data Services Performance and Optimization in 2G/3G

Edited by

G. Gmez and R. Snchez Both ofOptimi Corporation Spain

ffirs.fm Page iii Tuesday, March 22, 2005 7:26 AM

Copyright 2005 John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England

Telephone (+44) 1243 779777

Email (for orders and customer service enquiries): [email protected] Visit our Home Page on www.wiley.com

All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, or emailed to [email protected], or faxed to (+44) 1243 770620.

Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The Publisher is not associated with any product or vendor mentioned in this book.

This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought.

Other Wiley Editorial Offices

John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA

Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA

Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany

John Wiley & Sons Australia Ltd, 33 Park Road, Milton, Queensland 4064, Australia

John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 129809

John Wiley & Sons Canada Ltd, 22 Worcester Road, Etobicoke, Ontario, Canada M9W 1L1

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books.

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

ISBN-13 978-0-470-01180-5 (HB)ISBN-10 0-470-01180-7 (HB)

Typeset in 10/12pt Times by Integra Software Services Pvt. Ltd, Pondicherry, India.Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire. This book is printed on acid-free paper responsibly manufactured from sustainable forestry in which at least two trees are planted for each one used for paper production.

ffirs.fm Page iv Tuesday, March 22, 2005 7:26 AM

http://www.wiley.com

Contents

List of Contributors x

Foreword xii

Preface xiv

Acknowledgements xvii

1 Introduction 1John Cullen, Mattias Wahlqvist and Gerardo Gmez

1.1 Mobile Services in Perspective 11.2 Mobile Technology Evolution 3

1.2.1 Reasons for Mobile Technology Evolution 31.2.2 Mobile Technology Evolution Paths 4 1.2.3 Harmonization/Evolution Challenges 71.2.4 Future Outlook 8

1.3 Motivation for QoS 81.3.1 Service Experience 8 1.3.2 Radio Network Performance 10 1.3.3 Network Capacity 10 1.3.4 Network Design 10 1.3.5 Application Design 111.3.6 Service-Enhancing Technology 111.3.7 Conclusion 12

References 12

2 Cellular Wireless Technologies 13 Petteri Hakalin, Pablo Tapia, Juan Ramiro-Moreno, Raquel Rodrguez, Ma Carmen Aguayo-Torres and Rafael Snchez

2.1 Introduction 132.2 GSM/GPRS/EDGE 14

2.2.1 Description of the GSM System 15 2.2.2 The GSM Transition to Packet-Switched Systems (GPRS) 18 2.2.3 EDGE: The GSM Evolution 212.2.4 (E)GPRS Performance 21

ftoc.fm Page v Tuesday, March 22, 2005 7:27 AM

vi Contents

2.3 WCDMA/HSDPA 22 2.3.1 System Architecture and RRM 22 2.3.2 Transport Channels and their Mapping to the Physical Layer 242.3.3 Physical Layer and Air Interface 252.3.4 The HSDPA Concept 27

2.4 IS-95/CDMA2000-1x, EV-DV, EV-DO 28 2.4.1 CDMA2000-1x vs 3GPP UMTS 302.4.2 CDMA2000-1x Reference Architecture and QoS 302.4.3 Basic Voice Service with CDMA2000 33 2.4.4 Packet Data Operation with CDMA2000-1x 33 2.4.5 CDMA2000-1x Performance 362.4.6 Mobility 38

2.5 WLAN 392.5.1 Complementary WLAN Access Technology for

Cellular Networks 402.5.2 WLAN-3GPP and WLAN-3GPP2 Architecture 41

2.6 Future Outlook 442.6.1 Heterogeneous Networks 44 2.6.2 Physical and MAC Layers Trends 46

References 48

3 Data Services Architecture and Standardization 50Salvador Hierrezuelo, Alejandro Gil, Juan Guerrero, Raquel Rodrguez,Juan Torreblanca, Mattias Wahlqvist and Gerardo Gmez

3.1 Introduction 50 3.1.1 Circuit-Switched and Packet-Switched Services 50 3.1.2 Services Architectures and Protocols 51 3.1.3 Services Selection 51

3.2 Services Architecture 523.2.1 Services and Service Enablers 54 3.2.2 IP Multimedia Subsystem (IMS) 55

3.3 Data Protocols Characteristics 563.3.1 TCP/IP Networks 57 3.3.2 Impact of Radio Interface on Transport Protocols 67

3.4 SMS/MMS 683.4.1 Introduction to SMS 683.4.2 SMS Architecture and Signaling 69 3.4.3 SMS Protocol Stack 72 3.4.4 Introduction to Multimedia Messaging Service (MMS) 73 3.4.5 MMS Architecture and Signaling 733.4.6 MMS Protocol Stack 76

3.5 WAP 77 3.5.1 Introduction 773.5.2 WAP Architecture 783.5.3 Protocol Stack 793.5.4 Signaling 83

3.6 Web 85 3.6.1 Introduction 853.6.2 Architecture 85 3.6.3 Protocol Stack 86 3.6.4 Signaling 87

ftoc.fm Page vi Tuesday, March 22, 2005 7:27 AM

Contents vii

3.7 Push-to-Talk over Cellular (PoC) 88 3.7.1 Introduction 88 3.7.2 PoC Architecture 903.7.3 PoC Protocol Stack 93 3.7.4 PoC Signaling 95 3.7.5 PoC Performance Requirements 96

3.8 Network Gaming Services 98 3.8.1 Introduction 98 3.8.2 Network Requirements 99

References 101

4 Quality of Service Mechanisms 103Raquel Rodrguez, Daniel Fernndez, Hctor Montes, Salvador Hierrezuelo and Gerardo Gmez

4.1 What is Quality of Service? 103 4.1.1 QoS Definition 103 4.1.2 Need for QoS Differentiation 104 4.1.3 QoS Standardization 1054.1.4 Data Services Classification 106

4.2 IP-Based QoS 107 4.2.1 Motivation of IP QoS Mechanisms 108 4.2.2 QoS Paradigms 1094.2.3 IP-QoS Management in UMTS Networks 114 4.2.4 Traffic Handling Mechanisms 115

4.3 QoS Architecture in 3GPP and 3GPP2 117 4.3.1 End-to-End QoS Introduction 117 4.3.2 Evolution of QoS in 3GPP Releases 118 4.3.3 IP Multimedia Subsystem (IMS) 126 4.3.4 3GPP versus 3GPP2 in QoS 129

4.4 QoS Policy Management 1314.4.1 Motivation for QoS Policy Management 1324.4.2 History and Evolution 1324.4.3 IETF Policy Model 134 4.4.4 Policy Management in Mobile Networks 137

References 137

5 End-to-End Service Performance Analysis 139 Rafael Snchez, Gerardo Gmez, Pablo Ameigeiras, Jorge Navarro and Gabriel Ramos

5.1 Introduction 139 5.1.1 End-User Performance Analysis 140

5.2 Service Performance Characterization 1425.2.1 Characterization of End-User Performance 143

5.3 Data Link Effects 145 5.3.1 Data Link Effects in (E)GPRS 1465.3.2 Data Link Effects in WCDMA 149

5.4 Transport and Application Layer Effects 1565.4.1 TCP Performance 156 5.4.2 UDP Performance 1705.4.3 Application Layer Effects 171

ftoc.fm Page vii Tuesday, March 22, 2005 7:27 AM

viii Contents

5.5 Impact of Network Dimensioning in the Service Performance 1735.5.1 Dimensioning Example for (E)GPRS Services 174 5.5.2 Dimensioning Methodology 182

References 185

6 Service Performance Verification and Benchmarking 186 Rafael Snchez, Manuel Martnez, Salvador Hierrezuelo, Juan Guerrero and Juan Torreblanca

6.1 Introduction 1866.2 Key Performance Indicators 189

6.2.1 Network KPIs 190 6.2.2 Service-Based KPIs 199

6.3 Trial Methodology 201 6.3.1 Trial Phases 2036.3.2 Main Measurements 2106.3.3 Tools 212

6.4 Technology Benchmarking 2166.4.1 Introduction 2166.4.2 Traffic Generation 2176.4.3 Test Case Definition for Benchmarking 219 6.4.4 Benchmarking Result Analysis 2206.4.5 Network Performance with User Multiplexing 2306.4.6 Push-to-Talk over Cellular (PoC) 235

6.5 Performance Analysis Example 237 6.5.1 Service Differentiation Impact on Capacity and

Performance 238References 241

7 Customer Experience Management 243 Brian Carroll

7.1 Overview of Customer Experience Management 2437.1.1 The Challenge 245 7.1.2 The Solution 2457.1.3 Driving Mobile Revenue 246 7.1.4 Maximizing Operational Efficiency 247 7.1.5 Enhancing Customer Care 247 7.1.6 Measuring and Building Customer Satisfaction 248 7.1.7 Building a Brand Experience 2497.1.8 Improving Network Quality 249

7.2 CEM and Service Management 249 7.2.1 The Need for Service Management 249 7.2.2 The Service Management Landscape 251 7.2.3 Categorizing KQIs by Customer Experience 256 7.2.4 Architecture Options for Customer-Centric Service

Quality Management (SQM) 260 7.3 Advantages CEM Brings to an Operator 262 7.4 Summary 263References 263

ftoc.fm Page viii Tuesday, March 22, 2005 7:27 AM

Contents ix

8 Service Performance Optimization 264Gerardo Gmez, Juan Torreblanca and Mattias Wahlqvist

8.1 Introduction 264 8.2 Network-Level Optimization 266 8.3 Transport-Level Optimization 268

8.3.1 Standard TCP Recommendations from IETF 2698.3.2 Buffer Congestion Management 273 8.3.3 TCP Optimization in an Intermediate Node 2758.3.4 Datagram Congestion Control Protocol (DCCP) 277

8.4 Compression Techniques 278 8.4.1 General Fundamentals 2788.4.2 Content Compression Techniques 279 8.4.3 Wireless Specific Considerations 282

8.5 Performance Enhancing Proxies 284 8.5.1 Transport Layer Features 284 8.5.2 Application Layer Features 285 8.5.3 PEP Integration in Cellular Networks 286 8.5.4 Performance Improvement 287

References 288

Glossary 290

Index 296

ftoc.fm Page ix Tuesday, March 22, 2005 7:27 AM

List of Contributors

Ma Carmen Aguayo-Torres Dpto. Ingeniera de Comunicaciones Universidad de Mlaga, Spain

Pablo Ameigeiras Optimi Corporation, Spain

Brian Carroll Aran Technologies, Ireland

John Cullen O2, U.K.

Daniel Fernndez Optimi Corporation, Spain

Alejandro Gil Optimi Corporation, Spain

Gerardo GmezOptimi Corporation, Spain

Juan Guerrero Optimi Corporation, Spain

Petteri Hakalin Optimi Corporation, Spain

Salvador Hierrezuelo Optimi Corporation, Spain

Fergal Kelly (Foreword) Vodafone, Spain

fbetw.fm Page x Tuesday, March 22, 2005 7:26 AM

List of Contributors xi

Manuel Martnez Optimi Corporation, Spain

Hctor Montes Optimi Corporation, Spain

Jorge Navarro Optimi Corporation, Spain

Juan Ramiro-Moreno Optimi Corporation, Spain

Gabriel Ramos Optimi Corporation, Spain

Raquel Rodrguez Optimi Corporation, Spain

Rafael SnchezOptimi Corporation, Spain

Pablo Tapia Optimi Corporation, Spain

Juan Torreblanca Optimi Corporation, Spain

Mattias Wahlqvist Ericsson Espaa S.A., Spain

fbetw.fm Page xi Tuesday, March 22, 2005 7:26 AM

Foreword

The Customer is King is a phrase I believe many people will be familiar with and one that Ibelieve is highly relevant to the topic of Service Performance and Optimization. Brought backto its most fundamental driver, our networks exist because we have customers and it is thesecustomers that one way or another fund the development of our networks and our salaries!

As we have progressed from first generation systems through to third generation systems,things have become more sophisticated, largely due to the more ready availability of processingpower which in turn has led us to rely on more complex modulation schemes and greaterdetail into the transport protocols. A side effect of this complexity can be directly seen in thenumber of configurable parameters per cell, increasing with an approximate order of magnitudeacross each generation step. The consequence of this vast configurability is that serviceoptimization has become significantly more difficult.

Across the chapters of the book, the reader will be introduced to the various technologicalstandards and their corresponding architectures, including their in-built mechanisms forperformance management. Amongst these the opportunity for Quality of Service (often knownas QoS) differentiation is introduced. Whilst QoS introduces an autonomous management ofquality, the key word in my view is missing from its label differentiation, as differentiationis the fundamental leverage achieved from that approach.

Later on, the reader is introduced to the concept of end-to-end quality of service managementfrom an end-user perspective. For me this is good and bad. It is good in the sense that theend-user (also known as the customer) is considered albeit in aggregate form not individuallybut bad in the sense that the component-by-component approach whilst having some merit is,I believe, flawed.

In essence, the concept of component-by-component end-user end-to-end quality of servicemanagement is that, the system is made up of multiple components and that in order to managethe end-user quality, each of these must be assessed and compounded together to derive anoverall view on performance. Whilst this is theoretically correct, and certainly can be used asan approach, it is in my view an ageing approach. The reason I say ageing is that the numberof components that exist in a cellular network these days is huge and growing daily, so thetask of identifying them and all of the changes becomes almost impossible. Given this, it is inmy experience all too easy to omit components from the calculation of performance.

A complementary approach to try to overcome the main fallibility in the end-to-endapproach is now emerging. Leveraging again the improvements in processing power andrecognising that the system is an ever-evolving entity, the approach being taken at the leading

fbetw.fm Page xii Tuesday, March 22, 2005 7:26 AM

Foreword xiii

edge is now to sniff the interfaces between the key subsystems and to then correlate thetransactions with the chains of activity relating to a particular transaction of an individualcustomer.

This interface sniffing approach gives two valuable data sets. The first data set relates tothe concept of end-to-end quality of service management. By aggregating data by service-type, itis possible to see how the service is performing overall (i.e. the end-to-end view is achieved),as well as where precisely it may be failing. The second data set may be compiled by aggregatingthe data by individual customer. This approach is so far the only one that gives a truecustomer (end-user) perspective.

Having the true end-user perspective is a powerful tool. It allows us to do a number ofthings.

To talk to our customers about the quality they are individually experiencing, not a networkaverage that we have measured. Thus we improve credibility and customer satisfaction.

To establish priorities in issue management, perhaps choosing to address highest revenuecustomer issues first, maximising revenue.

To substantially increase revenues by identifying customer provisioning problems, releasingsuppressed revenue. This issue is generally not caught by the component-by-componentapproach and in my experience can easily add 5% to revenues, in spite of the best ofchecks and controls!

If ever there was a motivation to conduct service performance and optimization, surelythese must rank amongst the highest!

Fergal KellyCTO Vodafone, Spain

fbetw.fm Page xiii Tuesday, March 22, 2005 7:26 AM

Preface

Introduction

We are plunged into the era of mobile services. Wireless connectivity worldwide is becominga real need in advanced societies, not only for business, but also for entertainment, communitiesor even security purposes. In that sense, a huge variety of services is nowadays coexisting ina very complex and heterogeneous network infrastructure, which is additionally managed bydifferent parties. The end-to-end Quality of Service (QoS) is intended to achieve a seamlessintegration of the above-mentioned data services over the networks while providing the bestpossible experience to the end customers.

The complexity of monitoring and optimization processes of data service performance isquite evident, considering not only the quick diversification of emerging services and qualityrequirements associated with them, but also the coexistence of a huge variety of access technolo-gies and the wide coverage through which such technologies are offered.

The service performance optimization should not be just seen as a need for upgrading thenetwork with additional resources (over-provisioning), but to analyse the end-to-end scenario,service by service, in order to ensure a predefined service quality while minimizing the costs,i.e. optimize the network usage at the same time that customer experience is enhanced. Thisis the only way to minimize the unit cost of a call or data session.

The goal of a mobile network operator is to offer the customer an assured end-to-end QoS,with a variety of service levels and predictable service response. Such a goal requires a set ofintra-domain (radio access and core networks) and inter-domain agreements, which isconsistent along the end-to-end chain. For that purpose, it is very important for mobilenetwork operators and service providers to have the capability to measure the service per-formance as experienced by the customer, guarantee their QoS expectations and succeed inthe launch of new mobile services.

Do not expect to find here a magic parameter or formula that solves all the problems in thisrespect. This book is aimed at providing a proper methodology as well as the technicalbackground required for assessing and optimizing the End-to-End Service Performance overa multi-radio technology scope. A proper understanding of the key factors influencing theend-user performance is essential to fulfil such a goal.

We tried to approach these objectives by combining both theoretical background andanalysis with practical examples obtained from simulations and live network measure-ments, which could illustrate realistically what a user may expect from current cellularsystems.

fpref.fm Page xiv Tuesday, March 22, 2005 7:27 AM

Preface xv

Who Can Benefit from This Book?

Note that a very high number of entities and factors are involved in a mobile communication.Let us imagine a customer using his new 3G terminal for retrieving a real-time streamingvideo through the Web. In this simple scenario, an important group of experts in differentareas must contribute to optimize the customer experience (terminal manufacturers, contentproviders, mobile network operators, network element vendors, etc.).

This book is intended to cover many different aspects in this end-to-end approach focusingon the performance perspective. Therefore, we are sure that a wide group of readers couldbenefit from this book. First, mobile network operators (whatever the access technologythey are exploiting) may find useful guidelines and hints on how to assess, monitor,analyse and optimize the data service performance. Secondly, this book is also directedto mobile network vendors and developers, as well as telecommunications companiesworking on performance solutions, mobile applications and/or consulting services in thisarea. Finally, this book will also serve universities and other institutions with technicalbackground on telecom networks as the main reference on QoS and service performanceover wireless.

Content of the Book

The book is divided into 8 chapters, which can be summarized as follows. Chapter 1 introduces the quick evolution that wireless technologies and services have been

experiencing over the last years. This chapter also describes the motivation and the need forapplying QoS mechanisms in mobile networks as a way to satisfy end-user experience andoptimize network performance.

Chapter 2, Cellular Wireless Technologies, is aimed at providing an overview of currentand future radio technologies from a performance point of view. It describes the differentevolution paths and data capabilities to support emerging data services.

Chapter 3, Data Services Architecture and Standardization, describes mobile servicesarchitecture and functionalities, as well as introduces the reader to a particular set of services(including protocols, signaling and QoS requirements).

Chapter 4, Quality of Service Mechanisms, is aimed at providing an overview of whatQoS stands for, describing which kind of mechanisms are currently available in IP and cellularnetworks, introducing the need for QoS differentiation and providing a classification of dataservices according to their QoS requirements.

Chapter 5, End-to-End Service Performance Analysis, presents the concept of end-to-endperformance as the way to measure the quality of a network from the user point of view.Analysis methodology based on the cumulative effect of the different layers and networkelements through the transmission path is also analysed. The focus is placed on a technology-independent point of view that allows analysing the performance of the services based ongeneric parameters.

Chapter 6, Service Performance Verification and Benchmarking, provides a detailedview of the performance indicators and measurement methodology for assessing the qualityof a wireless data network. A combination of network indicators and service parameters pro-vides the best understanding of the system status. Service performance benchmarking resultsfrom live mobile networks are given and analysed in detail.

fpref.fm Page xv Tuesday, March 22, 2005 7:27 AM

xvi Preface

Chapter 7, Customer Experience Management, is intended to outline a new approach toservice management known as Customer Experience Management (CEM) or Customer Cen-tric Service Management (CCSM) a conceptual explanation of what CEM is; the reasonswhy CCSM is replacing traditional service management; and the impact on revenues andcustomer satisfaction will be covered in this chapter.

Chapter 8, Service Performance Optimization, is intended to provide some guidelines onhow to optimize the service performance, describing the main techniques available for thatpurpose and quantifying the performance gain for some of them.

Gerardo Gmez and Rafael SnchezThe Editors

fpref.fm Page xvi Tuesday, March 22, 2005 7:27 AM

Acknowledgements

First of all, we would like to thank all of the people that directly or indirectly have contributedto the publication of this book. We are really proud of the whole team that has allowed thischallenge to become a reality despite the great workload, full of tight schedules, trips andweekends of work.

Many thanks to Juan Melero for providing the opportunity to start this project and puttingus in contact with John Wiley & Sons, Ltd, in addition to his great advice whenever we needed it.Thanks also to Mark Keenan for his guidance and help to contact relevant people in the industrywhose input makes this book closer to operators and field engineers. It has been a real pleasureto include the contribution from Fergal Kelly as the Foreword.

This book would not have been possible without the contribution of all our good partners whogave their experience and support in many different areas: Pablo Tapia, Juan Ramiro-Moreno,Petteri Hakalin, Raquel Rodrguez, Juan Torreblanca, Salvador Hierrezuelo, Alejandro Gil,Juan Guerrero, Daniel Fernndez, Pablo Ameigeiras, Jorge Navarro, Gabriel Ramos, JohnCullen, Hctor Montes, Manuel Martnez, Ma Carmen Aguayo-Torres, Brian Carroll and JuliaMartnez. With special care, we would like to express our recognition to Mattias Wahlqvistfor his incredible help, support and planning capacity.

And of course, our immeasurable gratitude to all the colleagues who have supported, andeven paid theirselves, our measurements in the two continents and many different cities: JuanTorreblanca, Juan Guerrero, Juan Pablo Iriarte, Miguel ngel lvarez, Timothy Paul, HctorMontes, Greg Evans, Jay Langford and David Paolini. Special thanks to Salvador Hierrezuelowho not only contributed with many measurements and great analysis, but also took care ofall the technical preparations and software fine-tuning, which made it possible to perform allthe tests.

Not forgetting the great help from Optimi Corporation, allowing the usage of its softwareand always providing us with its full technical support and expertise in anything we needed,and all the operators and other companies we have worked with over the last years andwhich have provided us with the experience needed to face this challenge: Nokia, CingularWireless, AT&T, TIM, Telefonica, Vodafone, and many others.

Many thanks to the John Wiley & Sons, Ltd, publishing team (Mark Hammond, Sarah Hintonand Olivia Barnett) for their support during the development of this book.

This work is part of the research and development activities of Tartessos TechnologiesS.A. (Optimi Corporation). These activities are partially supported by Consejera de Empleo yDesarrollo Tecnolgico of the Junta de Andaluca.

flast.fm Page xvii Tuesday, March 22, 2005 7:27 AM

xviii Acknowledgements

Finally, we would like to thank our families and friends for their patience and love. I, Gerardo Gmez, would like to thank my parents, my brother and the rest of my family,

and especially express my loving thanks to Miriam. I, Rafael Snchez, would like to thank my parents and brothers who have always been

there. Thanks to my close friends who always push for a break those who are far away andnever forget and to the little baby-girl, Luca.

We welcome any comment or question you may have related to this book in order to con-tinue improving it. The e-mail address used for this purpose is: [email protected].

Gerardo Gmez and Rafael SnchezThe Editors

flast.fm Page xviii Tuesday, March 22, 2005 7:27 AM

End-to-End Quality of Service over Cellular Networks: Data Services Performance and Optimization in 2G/3GEdited by G. Gmez and R. Snchez 2005 John Wiley & Sons, Ltd

1 Introduction

John Cullen, Mattias Wahlqvist and Gerardo Gmez

1.1 Mobile Services in Perspective

Twenty years ago mobile phones were a rarity with less than 5 million subscribers world-wide. They tended to be fitted to cars as car phones as they were bulky and power hungry,used by the elite due to the high prices charged for equipment and service, provided onlyvoice call capabilities and only delivered service over what we would consider a small areatoday. At the same time, even those companies launching mobile services predicted that theoverall market would be very small. Ten years later, many industry observers still believedthat the market would remain relatively small.

Today, mobile devices are used by around 1.5 billion people worldwide, a three-hundred-foldincrease since 1985, which equates to a worldwide penetration slightly over 20%. Mobilecommunications is now a technology for everyone. For many people it is now an indispensablepart of their life with their mobile being among their key personal possessions alongside theirwatch and wallet.

The mobile device has changed all our lives and the way we live it. Listed below area number of examples.

Mobility: Today, we are travelling more for both business and leisure. This has lead toa heavier reliance on the mobile phone to stay in touch with colleagues and friends/family.

Planning: As everyone is reachable, we do not plan ahead. How many times have youheard: Yes. Lets meet at 12 in the city centre. Ill call you when I arrive, so that we findeach other.

Communities: Part of the tremendous popularity of mobile devices is that wherever youare, communication-wise you are very close to your friends and colleagues. Teenagerstoday rely heavily on their mobiles to keep in touch with their friends and to organize theirsocial lives. To do this they heavily rely on text to communicate with their community.

c01.fm Page 1 Tuesday, March 22, 2005 9:40 AM

2 End-to-End Quality of Service over Cellular Networks

Participation TV: TV shows are trying to appeal more to their audiences by allowing theiraudiences to interact with their shows so as to affect the outcome of the show (e.g. realityTV shows) or generate content (dating/chat shows) while also providing a revenue generationopportunity.

Marketing: Many consumer brands have started launching competitions or offers wherebyentries are made via SMS and an instant response can be given to customers. In somecases, prizes are downloads for handsets that allow customers to personalize their handsetswith ring tones or wallpapers. At the same time, the consumer brands are able to build upmarketing databases using entry information.

Security: Today, most of us would not think of travelling long distances in a car withouthaving a mobile with us in case of an emergency. Also today, in richer countries, manyparents are giving their young children mobile phones so that their children can contactthem in an emergency and so that they can keep track of their children.

For many young people today, their first commercial relationship with any communicationscompany is with their mobile operator. For this wireless generation, the mobile is at the hubof their social lives. When they first move away from home, they maintain a relationship withtheir mobile and in most cases the mobile service becomes the only communications servicethey subscribe to themselves. As a consequence, their relationship with a mobile operator istheir prime relationship with the communications industry replacing the traditional primerelationship enjoyed by fixed operators. Today, this unique relationship with the mobileindustry tends to be broken only when an individual moves into their own property and startsto consume services that require fixed lines such as broadband Internet services.

Looking forward, we are setting out as an industry on a new phase of market developmentwhere with Third Generation (3G) radio technologies the number of services and the richnessof those services is greatly expanded. Five years ago, the mobile industry talked about thehighest data rates that would be available for 3G. These high data rates are still an issue forlaptop PC users with data cards, but for average handset users 3G opens up the ability to usenew richer services and capacity that would not have been possible for mass-market customerswith Second Generation (2G) technologies. Listed below are a number of examples of howmobiles could be changing our lives in the future.

Communication: Video calling is starting to allow consumers to communicate face to faceand to share their environment with their colleagues. In todays busy world with frequenttravel, it allows families to keep in touch while on the move.

Entertainment: Music download and streaming is allowing people to get and listen tomusic on the move, releasing them from computers and fixed communications. At thesame time, the ability to download games, which is possible today, will be enhanced by thecapability to play them on the move with online friends so providing a new dimension togaming.

Current affairs: Already the first 3G operators are offering consumers the ability to keepup with events on the move via video clips so allowing consumers to be able to see, forexample, their team winning a sports game while on the move.

Content creation: The emergence of smart phones incorporating cameras, good qualitydisplays and reasonable processing capabilities will allow consumers to create and sharecontent. Content could be owner-generated pictures, videos, audio, text or any combination


Introduction 3

of these media types. Sharing could be by picture/video messaging, via online electronicjournals (blogs) or by peer-to-peer file sharing. To safeguard personal content, networkbackup capabilities will become essential.

Purchasing: The arrival of large colour displays on devices will make it more practical forconsumers to buy services from the Internet and carry out transactions on the move sofreeing consumers from their fixed PCs and allowing them to make use of dead time whentravelling, waiting for friends, etc. It will also provide a rich channel for governments tocommunicate and interact with their citizens.

Business: On our company networks today we have from our PCs high-speed access tocompany resources and to the Internet. On the move, our PC connectivity has been limitedby either connection speeds or the availability of hotspot coverage. The emergence of 3Gtechnologies will enable us to improve this situation by providing coverage over largegeographic areas.

Like the Internet world, the success of mobile data services will be built for giving con-sumers access to a rich set of services so as to satisfy a multitude of customer needs at theright cost. Unlike the Internet world, in the mobile environment the winning services andservice providers will be determined not only by the simplicity of using services on themove but also by the quality of the experience in using services the best service in theworld will not sell if a user needs an answer in ten seconds and it takes one hour. This bookaims to look at how the service performance can be tailored to give the right performanceat the right price.

1.2 Mobile Technology Evolution

Today, mobile telephony is a global industry with a global footprint in a large part of thepopulated world. In the beginning, however, mobile telephony systems were typically a localsolution on country level.

1.2.1 Reasons for Mobile Technology Evolution

There has been a tremendous evolvement of mobile telephony during the last 20 years, bothtechnology-wise and service-wise. One interesting aspect of the evolution of mobile technolo-gies is to ask yourself what is really the driving force being the engine for the switchoverfrom one technology to another. That is a complex question, and there is not one true answer.It is also so that the answer will depend on whom you ask. Here we anyhow try to illustratethe complexity of this question by giving a few opinions from different points of view.

Customer service requirements: Is it so that end-users are demanding better and morerequiring services, which leads operators and vendors to implement new technologies? Thisstatement is partly true and it is important to observe that it will likely become truer astime passes. In the beginning, the mobile telephone service was just a telephone serviceyou could use on the go. Today, there are additional services (SMS, WAP etc.) that areadding new requirements to the system. It is also so that the end-users today are muchmore advanced in terms of comparisons with, for example, services on the fixed Internet.



If a person can download a large email on the fixed computer, there should be no reasonwhy he/she should not be able to do it in his/her mobile phone.

Customer and traffic growth: Is it so that the growth in the customer base and the traffic thatgenerates are implying that the operators need to reinvest in newer more efficient systems?This is not really true. Typically, new features (e.g. half-rate codecs, frequency hopping etc.)are introduced to enhance capacity and quality, but it is of course important for the operator toprotect his CAPEX investment as long as possible. It is also so that the time to design a newsystem makes it impossible to rely on a new more spectrum efficient handling of the traffic.The problems are here today, and the future system will take many years to get into the field.

Differentiation of services and Quality of Service (QoS): Is it so that new systems aredeveloped to be able to perform service differentiation and offer QoS? To some extent yes.It is a common understanding that service differentiation and QoS is the only way to cost-efficiently offer a wide range of services. Still, the service differentiation has already beengradually introduced in todays systems, and so making service differentiation a mainreason for the development of new systems is only partly true.

Spectrum availability: When new spectrum is made available there is of course an urge tomake use of it in the best possible way. Spectrum is a scarce natural resource, and theintroduction of new more efficient systems is done easily if it is introduced together witha new spectrum band.

End-user requirements: The end customer has normally a firm opinion on whether he likesa service or not (like in this context normally means that he thinks that it is worth payingthe stipulated price for getting the service). That opinion heavily affects his usage of theservice. Still, considering the time it takes for a service to become a mass-market service,makes us believe that it is not end-user requirements that are driving the need for newsystems. The majority of the end-users are not advanced enough to know what they willneed in a five-year time frame.

Commercial aspects: There are of course commercial aspects that influence the willing-ness to introduce new systems into the markets. Vendors might want to protect or increasetheir market share; operators might want to create a high-end profile towards their endcustomers etc. Considering the time frame to introduce new systems, it is anyhow clearthat the commercial aspects are mainly considered on high strategic level.

To conclude, we can see that there is a variety of reasons for new mobile systems to be intro-duced, with the strongest ones being the need to make more and more efficient use of a limitednatural resource, the spectrum. On top of that, there are a multitude of other reasons to consider.

1.2.2 Mobile Technology Evolution Paths

Analog technologies were dominant in the cellular market up to 1997, when their globalmarket share was exceeded by that of 2G digital technologies. From that date, the GlobalSystem for Mobile communication (GSM) revolutionarily changed the way we look at andthink of mobile telephony. After its introduction we have seen a rapid evolution of services,technologies and performance. GSM technologys market share shows a sustained growthand today it has become the global 2G standard, deployed by more than 460 operators aroundthe world and accounting for more than 70% of the total number of cellular subscribers.


Introduction 5

General Packet Radio Service (GPRS) technology, developed as a Packet Switched (PS)extension of the GSM network, allowed high-speed access to IP-based services and at thesame time it provided an efficient use of the network resources. Some time later, EnhancedData for Global Evolution (EDGE) technology increased the radio data rates by includingsome enhancements in the modulation and coding schemes. (E)GPRS can be consideredas the convergence point between the Time Division Multiple Access (TDMA) developedin North America and GSM technologies, and is the foundation for the PS domain of the3G Universal Mobile Telecommunications System (UMTS).

Another parallel technology evolution path is the one coming from cdmaOne. Despite animportant growth during its first year of deployment, cdmaOnes (and its main successors:CDMA2000-based family) market share has stabilized around 15% of market share in 2004.Although a natural evolution from CDMA20001x would be the support of 1xEV-DO(1xEvolution, Data Optimized) and 1xEV-DV (1xEvolution, Data and Voice), many CDMAoperators are currently migrating towards GPRS and EDGE technologies as an alternativeoption (with the later integration of WCDMA). This last option is however dependent on thecellular-operators licensed bandwidth, since WCDMA technology is currently not supportedin the 800-MHz band, the future availability of dual mode cdmaOne/WCDMA terminals andthe integration effectiveness of the technologies.

Expected WCDMA launch is nowadays becoming a reality as the evolution path of 2Gtechnologies, being already supported over several markets around the world. The convergenceof 2G technologies towards the UMTS multi-radio 3G evolution path is clear. The entities,such as operators, global associations and standardization bodies, which are representing anddriving the evolution of three out of the four current most representative 2G technologies,have endorsed the UMTS multi-radio evolution path. Figure 1.1 summarizes the evolutionpaths associated with the existing 2G technologies.

The evolution of the mobile technologys market share and how these technologies aredistributed around the world is depicted in Figures 1.2 and 1.3, respectively [1].

Thanks to the evolution of the networks towards PS technologies, data services haveexperienced a huge increase in terms of data transmission capabilities, leading to an importantincrement in operator revenues. Currently, SMS and MMS are still the most profitable,

CDMA2000 1xEV-DV

EV

All IPSIPGERANHSDPAMobile IPv6

GSM

cdmaOne

TDMA

WCDMA

GSM/GPRS/EDGE

CDMA2000 1x

GSM/GPRS

3G UMTS multi-radio

2G

2000

First step to 3G

2001

3G Phase1Networks

20022003

Evolved 3G Networks

2004+

CDMA2000 1xEV-DO

Figure 1.1 Mobile technology evolution paths [2]



although other services like email, content downloading (i.e. Java applications, games, tones,etc.) or streaming are already pushing hard.

SMS and MMS, together with ring tones and information downloads, have representedbetween 2 and 7% of operators revenue in both North American and Latin American regionsin Q2 2004. China Mobile handled more SMS than any other operator, 30.9 billion in Q22004 [1]. Although in those regions, CDMA20001x was the most widely deployed technology,

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

CDMA

GSM

iDEN

PDC

US TDMA

W-CDMA

Analogue

Figure 1.2 Mobile technologys market share (forecasted from 2005 onwards) [1]

GPRS

EDGE

CDMA2000 1X

GPRS and CDMA2000 1X

EDGE and CDMA2000 1X Source: World Cellular Data Metrics, September 2004

Figure 1.3 Mobile technology deployment [1]


Introduction 7

data usage has been boosted by the continued deployment of advanced data networks (GPRS/EDGE), as shown in Figure 1.3.

In Europe, GPRS and partially WCDMA have been deployed until today, where an averagedata percentage of revenue reached 13.6% in Q2 2004. SMS traffic in western Europe grewapproximately 1718% in the 12 months by the end of June 2004 [1].

MMS had been launched commercially by 237 operators in 88 countries in September2004. MMS usage and traffic volumes on the whole remain low, being KTF Korea and Verizon(USA) the ones reporting a higher number of MMS (over 21 million in Q2 2004).

Total mobile subscribers to GPRS, CDMA2000-1x, I-mode and other advanced data servicesexceeded the 150 million mark in Q2 2004, and the total reached just over 152 million as at30 June 2004, or 9.9% of the worlds total mobile users. The reader is kindly referred toChapter 2 for a detailed description of the different technologies listed along this section.

1.2.3 Harmonization/Evolution Challenges

With the design and commercialization of a new system, there are a large amount of requirementsthat need to be considered not only from a technical perspective, but also from an economical andcommercial point of view. Here we list a few major challenges that are important to consider.

Backwards compatibility: In order to get maximum reuse of older investments, a naturalevolution also requires the new system to be backwards compatible towards older, alreadycommercially deployed systems. For example, with the introduction of UMTS in Europe, it isof high importance that already from day 1 being able to perform inter-system handovers to andfrom the commercial deployed GSM network. The reason for this is obvious, as the operatorwants to be able to offer continuous service coverage. Note, however, that general require-ments on backwards compatibility can mean many very different technical requirements onthe new system. An operator might want to make maximum use of his already deployednetwork, which might lead to, for example, requirements on inter-system handovers and theability to co-site the two systems. On the other hand, a mobile phone vendor might want toease the implementation of a multi-system handset, and might want to set requirements onclock frequencies to be used, as well as limiting the complexity between how the two systemsinteract.

Service transparency: New systems typically offer new services. For that reason it can bedifficult, especially in the beginning, for the operator to offer a continuous service supportover the whole coverage area. For example, high-speed data service is impossible to maintain inUMTS when the user leaves the UMTS coverage area and are handed over to GPRS. Fromthis aspect, it is anyhow considered important to be able to maintain some type of service,even if the service level is lower. Whether that is useful or not for the end-user is anyhowvery service dependent.

Interoperability/roaming: As users move between systems, it is difficult for the operatorto maintain a constant service level. In the same case as in the service transparency exampleabove, a user that is roaming into another network might not get all the services he/she cannormally access in his/her home network, as they might simply not be implemented.Another consequence is the implications this might have on the billing models to be used.



1.2.4 Future Outlook

It is clear that the development will not stop here and now. New systems and features are todaybeing standardized and developed for inclusion in the upcoming years. A global traffic growthtogether with the release of new spectrum or reforming of old spectrum will also increase theneed for newer more efficient ways to transport mobile data communication. Lately, a few trendshave however emerged that might have significant impact on how future mobile communicationsystems are designed and deployed, although they are not going to change the evolution path.

Emerging markets: There are still large parts of the world where mobile communicationhas not yet been deployed and there is a large market potential. Typical situation for thesecountries is that they have a high potential subscriber base, and that the fixed phone infra-structure is not so developed. What holds back a massive deployment in these countries isnormally that the average amount that can be charged to the subscribers is relatively low,which makes it challenging to deploy and market a network in a cost-productive way.Ultimately, this might lead to requirements to develop low-end systems with lowerproduction cost and less features.

New services: With the exponential growth of the Internet, there is also an explosiveamount of new services that users can access. By getting used to access these services onthe fixed Internet, there will be a demand for doing at least some of them while beingmobile. In addition, mobile communication can offer a multitude of mobile services thatwill also add new requirements on any future system.

New users: The introduction of mobile data services also opens up for a complete newsubscriber group machines. The vending machine can itself send its order for new drinksor to get service, or you could remotely find your car on a map and demobilize it when itgets stolen. This is an area becoming more and more important with an infinite amount ofpossibilities that we likely will see.

1.3 Motivation for QoS

The motivation to look at QoS is two-fold.

1. To provide a service experience to consumers that meets their expectations so that they aremore likely to use it again and recommend it to friends or colleagues.

2. To achieve optimum loading of an operators network so that the desired serviceexperience is delivered for each customer while maximizing network utilization.

The following section provides a brief introduction to the main factors involved inaddressing these two issues.

1.3.1 Service Experience

In the early data services market, many consumers are impressed to just use a service whenmobile. However, once this euphoria has passed, the vast majority of consumers start to judgea service based on how it meets their needs and expectations. As an example, the Short


Introduction 9

Message Service (SMS) was designed as a store and forward service, and was offered as suchby most operators in the mid-1990s with messages sometimes delayed hours before delivery.Consumers have conclusively taken to the service as a way of communicating quickly andefficiently with friends and colleagues without the need for a conversation. However, inmeeting these needs we expect that a message is delivered almost instantly. When importantbusiness and social meetings are arranged by SMS, a couple of hours delivery delay is unac-ceptable. As we can see from this example, a consumers expectations of a service dictatewhether it is perceived as working well or badly. These expectations in turn determine thecritical success factors that the network must deliver against if the service is to be perceivedas good. Take the following examples.

1. A customer using an always on email application (e.g. Blackberry) expects their emailsto be accurately received and that they are received within a reasonable time, for example1020 minutes after being sent. This implies that the network must deliver accurateinformation, i.e. a very low bit error rate, but that the payload can be delayed for areasonable amount of time.

2. A customer using a Push-to-Talk application will expect to get voice messages withina couple of seconds from their friend or colleague sending a voice message but must beprepared to tolerate some voice distortion on limited occasions. This implies that thenetwork must expedite the voice messages through the network but that limited packetloss can be tolerated.

3. A customer browsing the Internet from a laptop PC with a 3G data card will expect thatthe Web page loads accurately to a point where they can start reading in less than10 seconds, otherwise their concentration will lapse making the service uncomfortable touse. This implies that the network must deliver accurate information and that some limiteddelay can be tolerated. In this example, the way the Web page is built can also makea difference. For example, a Website that displays text within 10 seconds but thendownloads images in the next 10 seconds will often appear to be quicker than a page thatcompletely downloads in only 15 seconds.

In all these examples, if the network achieves the critical success factors, the consumeris likely to perceive a service as working well. If the network fails to deliver, the consumer islikely to perceive a service as working poorly.

When considering the quality of a network it is worth remembering that services run eitherbetween two terminals or a terminal and a server, and that the critical success factors applyacross the whole connection. There is no point, for example, in engineering the GPRSnetwork to meet the consumers expectations when the connection to the content provider isnot to the same standard and hence degrades the overall experience.

The main causes of a network failing to deliver against the critical success factors are:

Radio network performance Are there a lot of errors on the radio interface? Network capacity Is there sufficient capacity to deliver a good service? Network design Is there too much delay in the system; is sufficient capacity available end

to end? Application design Are the right protocols being used for a mobile environment? Service support Is service enhancement technology correctly configured?



1.3.2 Radio Network Performance

A well-planned radio network where data errors on the air interface are minimized in mostcases will improve application performance. If there are a large amount of errors, retrans-missions are required which can slow down the amount of information that can be transferredby protocols such as TCP. Radio errors also introduce extra delay into any conversationbetween application clients/servers, slowing down application response times. In a similarway, voice conversations can be slowed down by satellite delays.

1.3.3 Network Capacity

Well-designed mobile networks are dimensioned so that they have just enough capacityduring busy periods. Any more capacity than necessary adds network cost for operatorsreducing profit margins. Too little capacity and customer applications will not be able to getnetwork capacity so will deliver a poor customer experience. It is, however, extremely diffi-cult to predict loading accurately as demand may fluctuate by time, day, month and season aswell as demand growing with time. As a result, any operator wishing to offer customers agood experience, so as to encourage the uptake of services, would have to expensively overdimension their network to avoid congestion. To overcome this problem, standards havedefined the concept of QoS and this is starting to be implemented into network equipment.The QoS concept encapsulates the idea that different data streams could be treated differentlyby the network depending upon the service being carried (Chapter 4). Ideally, a service thatrequires fast response time is assigned a QoS that in periods of congestion it would receivepriority over other traffic. Conversely, a service that can tolerate a reasonable delay wouldhave lower priority than other traffic. By assigning different QoS to different services, whencongestion occurs traffic can effectively be smoothed over time with high priority traffic stillbeing transmitted with little delay but lower priority traffic being delayed until capacity isavailable. As a result, in periods of temporary congestion, a network providing QoS can meetcustomer-service expectations with existing capacity. In this way, with QoS, operators canmore effectively load their networks as they can tolerate temporary congestion while at thesame time ensuring that they deliver customers with the good service experience they expect.

It is worth noting, however, that QoS mechanisms are designed to work in periods oftemporary congestion, where lower priority traffic can be delayed without impacting theservice experience of those services. If heavy congestion occurs or the network is congested forextended periods of time, QoS cannot be relied on to maintain the customer-service experience. Inthese cases, further capacity is required and the network should be re-dimensioned to a levelwhere only temporary congestion occurs.

1.3.4 Network Design

The principle aspects of network design that can impact service performance are:

System delays Every additional piece of end-to-end delay slows down applicationconversations. For services where customers demand low response times, these delayscan make a service unusable if the application requires an extensive conversation.


End-to-End Quality of Service over Cellular...

Documents

Transcript of End-to-End Quality of Service over Cellular...